We've seen that, in the situation of a batter reaching third with two outs, that all runners score at the same rates for each of the subsequent events, and therefore that the variation in run scoring from that position reflects the variation in talent of the subsequent hitters.
What happens if we keep the outs constant, but advance the batter to a different base?
[more ]Having looked at runners on third with two outs, I turned my attention to the case of a runner on third with no outs. Clearly, we should still expect any hit to score the runner, but we have the additional possibilities of sacrifice hits and sacrifice flies to consider.
The information used here was obtained free of charge from and is copyrighted by Retrosheet. Interested parties may contact Retrosheet at 20 Sunset Rd., Newark, DE 19711.
When examining run scoring for this case, I found that the top two spots in the order were scoring significantly more often (87%) than the rest of the lineup (81%). What's different?
[more ]It seems reasonable to assume that, to first order, the chances of scoring from third with two outs won't depend a lot on the talent of the runner. Obviously, speed will matter on a passed ball or wild pitch, but we'll represent that these events are relatively infrequent, and can be dismissed in our early approximations, and can limit ourselves to the cases where the batter reaches base.
In this study, I'm tracking runners by the base they reached as a batter. So the cases I'm looking at are unlikely to be influenced by free passes (because the bases behind them are known to be initially empty, and it would take three passes to bring the runner home). So we can concentrate on the relationship between batting average and run scoring.
[more ]The data for this project comes from the event files provided by Retrosheet. You'll be seeing a lot of the following notice.
The information used here was obtained free of charge from and is copyrighted by Retrosheet. Interested parties may contact Retrosheet at 20 Sunset Rd., Newark, DE 19711.
Retrosheet offers game logs for the AL and NL from 1969-1992; these were all used. There are also logs available for earlier years in the AL, but I choose not to incorporate these because of the lack of a matching NL set. Thus, they remain viable for any who wish to test the conclusions of this study using independent data.
For each of the event files, it is a relatively simple matter to run the BEVENT utility to extract the desired information. The BEVENT output is saved into working files, and the working files processed into reports for each season. These can then be manipulated as desired.
[more ]How much does baserunning matter in run scoring?
It may seem that way, especially if one looks at runs scored as a function of times on base. Are the lumbering sluggers clogging the base paths, and failing to score as often as they should? Does this suggest that OBP overrates these players, crediting them for scoring more often than they do?
There are a few obvious sources of noise that should be considered when addressing the question. The probability of scoring is certain to be dependent on the number of outs, and not all runners will see the same distribution - leadoff hitters get more than their fair share of chances to reach base with no outs. Similarly, the probability of scoring must be dependent on the hitters behind a player in the lineup.
The purpose of the study is to get a sense of how large this noise is relative to the signal, to attempt to control for it, and to get some sense of how many runs a player should score, given the context he is playing in. From there, it may be possible to put an estimate on the scoring variation caused by running skills.